Entity Database for Entity-Triggering Question Answering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information retrieval systems on user devices fail to effectively identify and respond to entity-triggering queries, which require specific entity information, often providing irrelevant or incomplete answers due to the lack of precise entity identification and attribute matching.
Innovation Solution
A method is implemented where user devices classify questions as entity-triggering, obtain search results, identify and score relevant entities based on their association with documents, and output answers referencing entity attributes, using techniques such as similarity threshold matching and entity scoring to ensure relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional information retrieval systems are used to answer questions, then the system can provide general information, but the accuracy and relevance of answers to entity-triggering queries deteriorates
Solution Approach 1:
The system segments the question answering process into distinct stages: question classification to identify entity-triggering questions, entity identification from search results, entity scoring/ranking, and answer generation. This segmentation allows each stage to be optimized independently, improving overall answer accuracy for entity-related queries
Solution Approach 2:
The patent introduces an intermediary entity identification and scoring module between traditional search and answer generation. This intermediary component processes search results to extract, identify, and rank entities before generating answers, serving as a bridge that transforms general search capabilities into specialized entity-based answering
2Measurement precision
If the system processes more search results to identify entities, then entity identification accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The system employs partial action by scoring and ranking entities to identify the most relevant ones, rather than processing all possible entities equally. By focusing computational effort on the top-scoring entities that are most likely to be relevant, the system achieves high accuracy without processing excessive data
Solution Approach 2:
The patent changes the parameter of entity evaluation from binary (present/absent) to continuous (scoring/ranking). By assigning scores based on multiple factors and using thresholds, the system efficiently filters and prioritizes entities, reducing processing time while maintaining identification accuracy
Data Source
AI summary
An embodiment may receive a question at a computing device; obtain a search result set in response to the question; identify, using the computing device, one or more entities that are associated with at least one document referenced by the search result set; select, using the computing device, one or more relevant entities identified as being associated with (i) documents referenced by the search result set and (ii) the question; and output, using the computing device, an answer to the question based at least on the selected one or more entities.